[mike777]

Yes, I did assume all of these factors you point out Todd are being improved at a cost/performance exponential rate equal to or greater than cpu speeds. That includes software performance/efficiency/productivity, however that is measured and compared.

If the answer is no I am surprised to say the least.

[hrothgar]

helene_t, on Feb 15 2007, 01:34 AM, said:

Same thing. The bottlenecks of the internet are the hups. Faster chips means more bandwidth.

This is a drastic over-generalization.

Here in the US, the main bottle neck is whats call the "last mile".
Connectivity from a local home to the backbone is dreadful

[jtfanclub]

DrTodd13, on Feb 14 2007, 05:48 PM, said:

The size of memory has gotten better faster than its latency (speed of light problems yo) but still not at an exponential rate. Memory bandwidth has gotten better but not exponentially with higher front-side bus rates and multiple memory ports.

Storage - very similar story here. The disk is mechanical and the head can only move so fast so latency is terrible. We are talking roughly 10ms which is an eternity in CPU cycles. Again, capacity and bandwidth have gone up but not exponentially.

Network - latency...speed of light man. Capacity and bandwidth the same thing for networks and again they've increased but not at the same rates as CPU performance.

I don't know what you mean by software. Programmer productivity is probably what has improved the least over the years.

http://news.yahoo.com/s/pcworld/20070214/t...joRjrdWfdkjtBAF

Disclaimer: I do not speak for IBM, nor do I work in the department that's developing this.

IBM is going to double its on-board cache in the next year. With inherent parallelism (not written into the program), that will effectively almost double throughput on the L1 cache angle.

As far as disk drives go, this is just an expense thing. I support systems with over 100GB of RAM. With sufficient main RAM, and lots of disk cache, disk drive speed only really affects reboot time.

Computers are still growing in speed exponentially. If you don't believe me, buy a PS3. It makes the supercomputers of 5 years ago look like they're crawling (and in fact, people are building supercomputers out of PS3 chips).

[mike777]

As I recall another article said the goal was to make home pc's at today's prices the equivalent of supercomputers from 1995 by 2011 or 2012.

[paulg]

jtfanclub, on Feb 15 2007, 12:18 AM, said:

Computers are still growing in speed exponentially.  If you don't believe me, buy a PS3.  It makes the supercomputers of 5 years ago look like they're crawling (and in fact, people are building supercomputers out of PS3 chips).

Historically the development of supercomputing has been driven by the defen{c/s}e industry.

Now the fastest floating-point processors are found in graphics/video cards and it is the gaming industry that is the major force. And this means that supercomputing should be a lot cheaper since there is true volume.

However, as I believe Seymour Cray said, it is all about memory bandwidth (and not processor speed). Even more true today than in his day.

[mike777]

".....However, as I believe Seymour Cray said, it is all about memory bandwidth (and not processor speed). Even more true today than in his day. ..."



Another computer problem with a billion dollar solution?

[paulg]

DrTodd13, on Feb 14 2007, 11:33 PM, said:

There are essentially two kinds of programs...embarrasingly parallel and all the rest.  Things like raytracers fall into the embarassingly parallel group.  You can divide up the picture into chunks and each processor works on its own chunk and the processors don't have to talk to each other or exchange data much to do their work.  The existing programming model doesn't make it easy to write these sorts of programs but due to the limited interaction between processors it is not all that difficult either.  The current programming model includes C, C++, Java, .NET.  There is nothing special about Java or .NET that fundamentally makes it a better environment for parallel programming.  Conversely, there are a range of other programming models all of which you've probably never heard of that do make it easy to write such programs.  However, these other programs models only work on certain types of problems and it is not useful for general purpose parallel programming.

I fully agree with this in the HPC context.

However if you are writing an internet-facing application then the Java 2 and .Net environments make it dramatically easier to write a program that services thousands/millions of users.

Of course you may feel this is not real parallel programming

[paulg]

mike777, on Feb 15 2007, 12:22 AM, said:

As I recall another article said the goal was to make home pc's at today's prices the equivalent of supercomputers from 1995 by 2011 or 2012.

Essentially already happened.

Systems I helped sell in the mid 1990s for $10M are being replaced today by a $2500 server that is more powerful, has more memory and memory bandwidth, more I/O and consumes a fraction of the power.

But the demands of the home user are different from business. Although the processors may be the same, the system infrastructure is cost-constrained and there is more focus on the user interface (aka graphics card).

[helene_t]

hrothgar, on Feb 15 2007, 01:00 AM, said:

helene_t, on Feb 15 2007, 01:34 AM, said:

Same thing. The bottlenecks of the internet are the hups. Faster chips means more bandwidth.

This is a drastic over-generalization.

Here in the US, the main bottle neck is whats call the "last mile".
Connectivity from a local home to the backbone is dreadful

But this is still a chip performance issue, isn't it? I suppose the bottleneck is not the cable from your home to your ISP, but rather your network adapter or (more likely) some hardware at your ISP's place.

[hotShot]

helene_t, on Feb 15 2007, 01:05 PM, said:

But this is still a chip performance issue, isn't it? I suppose the bottleneck is not the cable from your home to your ISP, but rather your network adapter or (more likely) some hardware at your ISP's place.

Sorry Helene it is a cable problem. A shielding problem to be more precise. Assume you have a 2 wire cable to your place, than the signal will run all the way from ISP to you and back. The signal going back will have an effect on the wire leading to you, so the signal is distorted. This effect grows by cable length and signal speed. The solution to this is using shielded cables. Unfortunately telephone cables needed much less shielding than a broadband data connection.

[mike777]

So wireless is not the inexpensive solution for the last mile problem and for internet on our tv's?

[glen]

DrTodd13, on Feb 14 2007, 11:33 PM, said:

There is nothing special about Java or .NET that fundamentally makes it a better environment for parallel programming.

See the multi-threaded features in .NET.

cardsharp, on Feb 15 2007, 04:35 AM, said:

.Net environments make it dramatically easier to write a program

Of course .NET "easier" can be considered to be using a C-5 Galaxy instead of car to move stuff to Vegas. The aircraft would be faster and able to handle a lot more, but driving that C-5 is not really easier than the car, and does take a fair bit of fuel.

[hotShot]

Wireless networks need hotspots, and they make sense in populated areas.
Unfortunately while they work well, if you are outside, you'll have problems entering a building that has a steel frame or concrete walls. They are reducing the signal strength and cause data errors, so the data rate is reduced. Seems to me that this type of building has very common, in populated areas.
Areas where buildings are more transparent for EM-radiation, have usually a low population rate and hotspots don't pay of.

The bandwidth problem will be solved, TV on demand, VideoChat and other services you can make money with require more bandwidth. Most of the backbones are ready for it and the "last mile" will follow.
Soon all phone call will be voice-over-ip calls, at least least among the phone companies, this includes mobile phones.
You won't have different networks like phone, mobile phone, cable tv and Internet all these will unite to one network and you might have more than one device to access it.

[hotShot]

Instead of asking what effect technical developments will have on BBO, we should ask ourselves what features we would like to have using the BBO environment.

Lets take the 3d desktop environment, no one needs it, it requires a lot of powerful hardware for doing nothing, but is't looking great.

Would we like to have an avatar moving to a virtual tourney room, and sit at virtual tables?

We could wish for a BBO-Hardware. It gives us real card with a RFID chip and has a display on the table, showing the bids made or card played. When we put a card on this device, the play is transmitted to BBO. This won't need more processor power, just some electronik work and USB connector and a device-driver.
We could wish for voice commands, upps maybe vista can do that already.

We could wish for a "TD robot" who runs tourneys.
How about duplicate tourneys with GIB partners?

The BBO environment has lots of areas where tasks can be distributed to more than one computer, i bet it is already done. These areas don't need to wait for faster processors, one can just add another computer.
Other areas can't work parallel, the processing power may be a bottleneck there, but if all other tasks operate on different CPU's, even the existing hardware should have some reserve.

[hrothgar]

Its very important to differentiate between technology and business model. Just because a new technology liek WiFi comes along doesn't mean that its possible to create a viable business model to develop / deploy it.

Right now, the United States has just about the worst Internet service in the developed world. If you compare what we have available with countries like Korea, Japan, and the like we're a laughing stock. From my perspective, the difference boils down to one (simple) issue: In the US private phone / cable companies are responsible for developing / selling the network infrastructure. In most parts of the world, the government has taken a much more active role in provisioning this service.

I don't want to get into a long, involved debate about the relative merits of central planning versus market economies. I (hope) that anyone reading this will recognize that that large bureaucracies are large bureaucracies. AT+T and Comcast aren't any closer to a market driven economy than Gosplan or any of the old planning agencies in the Soviet Union. The only real difference is that AT+T'd mission in life is squeezing every nickle it can out of its customers...

I'll note in passing, that many of the most promising WiFi deployment models were predicated arround the concept of municiple wireless. Individual cities wanted to be able to own/operate their own WiFi networks. AT+T and COmcast was able to get a large number of state legislatures to pass laws banning muni wifi... Quite annoying.

[hrothgar]

hotShot, on Feb 15 2007, 04:36 PM, said:

Instead of asking what effect technical developments will have on BBO, we should ask ourselves what features we would like to have using the BBO environment.

I want to echo a point that I made in my last posting: Technology does not always equate to viable "business model".

As I have noted in the past, I firmly believe that major bridge events like the Bermuda Bowl, the Vanderbilt, and the Cavendish should be conducted using an electronic playing environment. I won't bother restate my arguments in full. I will simple note my belief that this would provide superior service for

1. Spectators (unrivaled opportunities for vugraph)
2. Players (dramatically improved security / elimination of many forms of UI)
3. Administrators (Easier to manager, though it would require a different skill set)

The technology that we need is available today:

Yes additional code would need to be written. Yes rigourous testing would be needed. However, to my knowledge, there is no work going on to support this type of project. All too often political intertia trumps new technology.

[glen]

hrothgar, on Feb 15 2007, 08:57 AM, said:

All too often political intertia trumps new technology.

See electronic poker tables (i.e. tables to be used for live play - for example see: Detroit Free Press: MotorCity bets on dealer-less electronic poker). Sometimes money trumps intertia.

[DrTodd13]

officeglen, on Feb 15 2007, 05:06 AM, said:

DrTodd13, on Feb 14 2007, 11:33 PM, said:

There is nothing special about Java or .NET that fundamentally makes it a better environment for parallel programming.

See the multi-threaded features in .NET.

cardsharp, on Feb 15 2007, 04:35 AM, said:

.Net environments make it dramatically easier to write a program

Of course .NET "easier" can be considered to be using a C-5 Galaxy instead of car to move stuff to Vegas. The aircraft would be faster and able to handle a lot more, but driving that C-5 is not really easier than the car, and does take a fair bit of fuel.

This is what I do for a living so I know what I'm talking about. Threads and locks suck. If this is your model for parallel programming you are going to have no end of problems. There's quite a bit of research showing that any parallel model based on threads and locks will have tons of inherent problems. You can try to abstract it away in cases but the problems will always be there.

[fred]

You will all get a taste of the future of BBO soon.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[Al_U_Card]

fred, on Feb 15 2007, 02:46 PM, said:

You will all get a taste of the future of BBO soon.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

Oh you tease! lol


On second thought....should we be worried????